Re: Dons flame resistant (3 hours) interface about Linked Data URIs

OK, lets take a step back from the discussion of redirection, and remind 
ourselves what we are trying to achieve ('cos we're all agreed on that, 
aren't we ;-) ).  Just in case that is an issue in itself, here's my 
list, in order of priority:

1. unique and persistent URIs are routinely used to represent subjects 
of discourse.  These are published, and are used by many people to 
represent said subjects of discourse in many machine-processible 

2. each such URI can be resolved to a machine-processible resource which 
provides sufficient information to define the subject of discourse and 
distinguish it from other subjects;

3. each such URI can be resolved to a human-readable resource which 
provides sufficient information to define the subject of discourse

These requirements only apply when minting Linked Data URIs.  You'll 
notice that they don't mention RDF ...

One thing I discovered when I implemented 303 redirects is that they 
weren't good enough for a popular Linked Data browser I was using to 
test my work.  Instead of requesting application/rdf+xml and getting the 
RDF directly, this browser was issuing a standard HTTP request, getting 
redirected to the HTML page, and expecting to be told by that page where 
the RDF lived.  To be precise, it was expecting to find a line like:

<link rel="alternate" type="application/rdf+xml" 
title="RDF" />

pointing back to the RDF which it _could_ have requested directly.

So, in cases where it is "too hard" to support redirection, I would 
suggest that we use this sort of linkage: the web page contains a <link 
rel="alternate"> which points to the RDF, and the RDF contains a 
statement which points back to the human-readable web page.  These two 
resources will of course have different URLs (since it's too hard to 
support redirection), so the final decision is: which of these URLs 
should be the one which represents the subject of discourse?  And the 
answer has to be: the one which points to the machine-processible 
resource (all right, RDF), since this is the NIR.

This would work, wouldn't it?


In message 
<EMEW3|5be2c6a51e97d74c7b86468f22901f20l691MZ02hg||C296%hg>, Hugh Glaser <> writes
>I am finding the current discussion really difficult.
>Those who do not learn from history are condemned to repeat it.
>As an example:
>In the 1980s there were a load of hypertext systems that required the users
>to do a bunch of stuff to buy into them. They had great theoretical bases,
>and their proponents had unassailable arguments as to why their way of doing
>things was right. And they really were unassailable - they were right.
>They essentially died.
>The web came along - I could publish a bunch of html pages about whatever I
>wanted, simply by putting them in some directory somewhere that I had access
>to (name told to me by my sysprog guru), and suddenly I was "on the web". If
>the html syntax was wrong it was the browser's problem - don't come back and
>tell me I did wrong, make what sense of it you can, it's your problem.
>Such simplicity, which was understandable by a huge swathe of people who
>were using computers, and acceptable to their support staff, simply swept
>all before it (including WAIS, ftp, gopher).
>Arguments about how "broken" the model was because of things like links
>breaking and security problems were just ignored, and now seem almost
>archaic to most of us.
>I want the same for the Semantic Web/Linked Data.
>Discussions of 303 and hash just don't cut the mustard in comparison. So I
>find it hard to engage in an extended discussion about them.
>Q: "How do I do x?"
>Me: "Try this."
>Q: "This doesn't work, what now?"
>Immediately says to me that "this" must be wrong - we should go away and
>think of something better.
>So would it really be so bad if people just started putting documents with
>RDF in on the web, where the URI for both the document and the thing it was
>about (NIR) got confused?
>All I actually want is a URI that resolves to some RDF.
>And even perhaps people would not run off to RDFa so quickly?
>If I can't simply publish some RDF about something like my dog, by
>publishing a file of triples that say what I want at my standard web site,
>we have broken the system.
><3 hours flame resistance starts />
>No virus found in this incoming message.
>Checked by AVG -
>Version: 8.5.375 / Virus Database: 270.13.9/2228 - Release Date: 
>07/09/09 18:07:00

Richard Light

Received on Friday, 10 July 2009 09:13:13 UTC